Provider Charges And State Surprise Billing Laws: Evidence From New York And California

Surprise medical bills, when patients are billed for out-of-network care they could not reasonably have anticipated or prevented, have drawn considerable attention in recent years.^1–3 During the past decade many states have passed laws that protect patients from such surprise billing and provide a method to determine the amount that insurers must pay out-of-network providers during these scenarios. When the federal No Surprises Act went into effect in January 2022, all patients nationally became protected from surprise bills. However, despite the fact that the No Surprises Act provides a process to adjudicate payment disputes between insurers and out-of-network providers, prohibiting arbitrators from considering providers’ billed charges in these scenarios, states with existing laws can continue to allow the consideration of charges. This difference in the way these insurer-provider disputes are adjudicated may have important consequences for overall health care costs and the financial burden shouldered by consumers.

Under state surprise billing laws, out-of-network provider payments for surprise bills are typically determined either by a payment standard (for example, the greater of 125 percent of Medicare prices or the average contracted rate for the health plan in that region) or by an independent dispute resolution process, depending on the state. Prior research shows that when independent dispute resolution processes include the consideration of physician charges by service and geographic region (for example, usual, customary, and reasonable charges from FAIR Health or a similar database), the result is payments to providers tied to the eightieth percentile of charges in that area.^4,5 For example, in New Jersey the median payment from an arbitration decision was 5.7 times the median in-network rate for the same services in 2019.⁴ It is possible that providers in states with independent dispute resolution processes where arbitrators can consider billed charges will increase their charges either to raise the arbitration benchmark value for the eightieth percentile of charges or to increase out-of-network reimbursement if their original charges were less than the eightieth percentile of charges. Ultimately, this may raise payments to providers for out-of-network care that goes to arbitration.

In this study we compared trends over time in provider charges for out-of-network care during surprise bill scenarios in New York, where the state law uses an independent dispute resolution process pegged to charges to determine provider payments, with charges in a comparison group of states where no surprise billing law exists. We also compared provider charges in California, where the state law uses a payment standard tied to in-network prices to determine provider payments, with charges in the comparison group of states.

Study Data And Methods

Data Source And Population

We used medical claims data for commercial health plan members from episodes between July 2011 and March 2020 from Elevance Health (formerly Anthem, Inc.), a large national insurer. The analysis included a random sample of more than 3.5 million commercially insured members who lived either in New York or California (states with surprise billing laws) or in Kentucky, Ohio, Wisconsin, Indiana, Georgia, Virginia, or Colorado (collectively the comparison group) and who had at least one nonemergency inpatient hospitalization during the study period. The comparison-group states are those served by Elevance Health that had not passed surprise billing laws by the end of 2019.

Episode Identification

Although some of the previous research on surprise billing focused on emergency department (ED) visits,⁶ our analysis focused instead on nonemergency inpatient hospitalizations for a number of reasons. First, we wanted to allow for a larger range of potentially infrequently performed procedures, for which providers may have greater leverage in affecting usual, customary, and reasonable charge benchmarks than for commonly performed ones. We also limited our analysis to nonemergency inpatient hospitalizations because California members were already protected from surprise bills during emergency care before the passage of that state’s surprise billing law. Furthermore, surprise bill scenarios during nonemergency care occur in in-network facilities with an in-network primary provider—a circumstance in which a patient is unlikely to anticipate care from an out-of-network provider.

We restricted our analytic data set to nonemergency inpatient hospitalizations occurring in the member’s state of residence. We defined such hospitalizations as those that did not include a revenue code for the ED (that is, the patient was not admitted through the ED) and for which there was no claim for an emergency ambulance on the admission date. Next we identified the subset of episodes that contained a surprise bill scenario, using a definition that closely mirrors the taxonomy for surprise bills during nonemergency hospitalizations presented by Christopher Garmon and Benjamin Chartock:¹ Nonemergency hospitalization occurred in an in-network facility, the primary provider was in network, and there was a claim by another provider for out-of-network care.

The primary provider was defined as the provider with the highest charged amount during the episode, although certain specialties including radiologists, pathologists, anesthesiologists, assistant surgeons, and surgical assistants were categorically excluded from being the primary provider. All episodes that met these criteria were included in the analysis, regardless of the charge or allowed amounts (see online appendix section 1 for further information on episodes where the charge and allowed amounts were equal).⁷

Surprise bill scenarios were identified regardless of whether the patient was responsible for a balance bill arising from the scenario, which depended on whether they were protected by a state law as well as the provider’s decision to balance bill if the patient was not protected.

Primary Outcome And Empirical Methodology

For each surprise bill scenario, we calculated the provider charges for all out-of-network care occurring during the episode of care. We compared the trends in average provider charges over time in New York versus the comparison states and in California versus the comparison states. The same states were used as a comparison for both New York and California, although comparison episodes occurring between the passage of New York’s and California’s laws were considered to be in the pre period for California and the post period for New York.

Using a difference-in-differences approach, we compared the changes in total out-of-network provider charges at the episode level (winsorized at the 1 percent level) from the pre- to post-law periods, adjusting for patient characteristics (age, sex, baseline comorbidity score, and fully insured or administrative services only plans), hospital characteristics (rural or urban, teaching hospital, and critical access hospitals), specialty of the out-of-network provider or providers, number of out-of-network claims in the episode, quarterly fixed effects, hospital fixed effects, and Current Procedural Terminology (CPT) code fixed effects for the highest out-of-network charged CPT code in the episode. We also used a provider-level specification to isolate the change in charges among providers who were out of network both before and after the law versus those who changed network status during the study period. We considered the quarter of the passage of the New York (2014 first quarter [Q1]) and California (2016 Q3) laws to define the pre and post periods. Detail on the regression specifications is in appendix section 2.⁷

Our analysis also included event studies to test the assumption of parallel pre trends; permutation tests in the spirit of Fisher to account for potential biased standard errors given a limited number of treated clusters; a specification at the hospital level rather than episode level; and an alternative pre and post period based on the effective date, rather than passage date, of the laws. As Zack Cooper and colleagues note, it is difficult to determine the exact date on which the law may have begun to affect charges.⁶

To further understand the patterns we observed, we calculated the changes in average unadjusted out-of-network charges by selected provider specialties in the pre and post periods in New York and California (see appendix section 3).⁷ Because assistant surgeons and surgical assistants had the greatest unadjusted increase in charges in New York during the study period and represented a substantial share (19 percent) of surprise bill scenarios, we repeated our main difference-in-differences regressions, event studies, and permutation tests while limiting the sample to episodes containing an out-of-network claim from an assistant surgeon or surgical assistant and limiting the out-of-network charges to those billed by assistant surgeons and surgical assistants. We excluded Wisconsin from these analyses because we observed very few surprise bill scenarios with out-of-network charges from an assistant surgeon or surgical assistant in Wisconsin during the study period.

Limitations

Our study had several limitations. In estimating the difference-in-differences models, our comparison group of states not treated by a surprise billing consumer protection law only comprised the states that are served by Elevance Health commercial plans. We therefore can assess the parallel trends assumption for provider charges in only the treated states (New York and California) and the seven comparison-group states that had no surprise billing patient protection. Without data from other states not served by Elevance Health, we could not verify that the parallel trends that we observed held more broadly in other states. However, parallel trends in a long pre period in the states we did observe lend support to our findings.

Second, standard errors may have been biased downward, as there were a small number of treated clusters (states) in our difference-in-differences regression, which may have increased the likelihood of false-positive findings. We implemented permutation tests to address this limitation.

Third, our analysis was limited to measuring charges for out-of-network nonemergency hospital care during surprise bill scenarios and might not be generalizable to other out-of-network care or provider charges more generally.

Fourth, although the data we used comprised claims from a large national insurer, our results might not be generalizable to the entire commercial market, where providers may contract with different health plans.

Finally, although we were able to identify changes in charges occurring after the state laws were passed, we were unable to definitively ascribe the changes to intentional or strategic provider behaviors.

Study Results

The sample of surprise bill scenarios by state in the pre- and post-law periods used in the analysis is described in exhibit 1. Overall, our sample included 28,245 surprise bill scenarios from New York (11.3 percent of episodes), 31,718 surprise bill scenarios from California (11.6 percent of episodes), and 60,810 surprise bill scenarios from the comparison group of states (5.7 percent of episodes). Among these surprise bill scenarios, 5,288, 6,642, and 14,754 from New York, California, and comparison states, respectively, contained an out-of-network charge from an assistant surgeon or surgical assistant (data not shown). Unadjusted trends in total and assistant surgeon/surgical assistant out-of-network charges in New York and California are shown in appendix section 4.⁷

We observed a difference in patterns of out-of-network inpatient hospitalization charges between New York and California, respectively, and the comparison group of states. Although trends were similar in both states compared with the comparison group before the passage of the law, charges in New York rose after the passage of the law. Overall, the adjusted change in charges from the pre to the post period was $1,157 greater in New York than in the comparison states ($p=0.006$) (exhibit 2, model 1). Given the pre-period average of $4,864 in New York, this represents a 24 percent increase in charges above trend. We present the event study results for New York in exhibit 3. In California, conversely, after similar baseline trends, we detected a $752 decrease in charges relative to comparison states ($p=0.011$) (exhibit 2, model 3). Given the pre-period average of $3,038 in California, this represents a 25 percent decrease in charges below trend. We present the event study results for California in exhibit 4.

SOURCE Authors’ analysis of claims data from Elevance Health (formerly Anthem, Inc.). NOTES This event study follows the specification of regression model 1 in exhibit 2. Quarter −1, the reference point, represents the last quarter in the pre period (2014 Q1). Effect size estimates the effect of the quarter on differences in out-of-network-charges between New York and comparison states relative to quarter −1.

SOURCE Authors’ analysis of claims data from Elevance Health (formerly Anthem, Inc.). NOTES This event study follows the specification of regression model 3 in exhibit 2. Quarter −1, the reference point, represents the last quarter in the pre period (2016 Q3). Effect size estimates the effect of the quarter on differences in out-of-network-charges between California and comparison states relative to quarter −1.

Provider-level panel regressions taking the mean charge per provider that maintain the composition of out-of-network providers in the pre and post periods also resulted in an increase in total out-of-network charges in New York ($815, 10 percent; $p=0.002$) and a decrease in California ($474, 15 percent; $p=0.001$) (exhibit 2, models 2 and 4 for New York and California, respectively).

We observed a large increase ($4,358; $p<0.001$) in assistant surgeon/surgical assistant out-of-network charges in New York compared with other states (exhibit 2, model 5), representing a 43 percent increase over baseline. When we estimated the within-provider model (exhibit 2, model 6) in New York, we observed a $3,802 increase over baseline (29 percent). The change in charges in New York seems to begin about a year after the passage of the law, which is when the law went into effect (see appendix section 5).⁷ We did not observe significant changes in assistant surgeon/surgical assistant out-of-network charges in California compared with other states (exhibit 2, models 7 and 8; see appendix section 5 for event studies showing trends over time).⁷ Permutation tests show that New York had the largest increase in comparison with the alternative states when they were counterfactually assumed to have been treated states for both total and assistant surgeon/surgical assistant charges. Permutation tests also show that California had the largest decrease for total charges but was in the middle for assistant surgeon/surgical assistant charges. Figures in the spirit of Fisher permutation tests are shown in appendix section 6.⁷

Results of sensitivity analyses are in the appendix.⁷ Hospital-level panel regressions taking the mean charge per hospital resulted in similar estimates, with a $1,009 increase in total out-of-network charges in New York ($p=0.046$) and a $900 decrease in California ($p=0.001$) (see further detail in appendix section 7).⁷ Finally, using the law’s effective date instead of its passage date resulted in similar estimates, with a $1,157 increase in New York ($p=0.003$) and a $929 decrease in California ($p=0.009$) (see further detail in appendix section 8).⁷

Discussion

Using a data set of commercial insurance claims from Elevance Health, a large national insurer, we found that in New York, provider charges increased for surprise bill scenarios that arose from inpatient nonemergency hospitalizations after the passage of a surprise billing law that uses an independent dispute resolution process relying on charges to determine out-of-network provider payments. In contrast, we found a decrease in provider charges for a parallel set of episodes in California, whose law instead relies on a payment standard tied to in-network prices to determine out-of-network payments during surprise bill scenarios. These findings were robust to several regression specifications. The magnitude of the difference decreased slightly in New York (from $1,157 to $815) in the provider-level panel regression, suggesting that some of the total increase in charges may be due to changes in the composition of out-of-network providers. However, we still detected a significant increase in charges among providers with out-of-network claims in both the pre and post periods relative to the comparison.

We are unaware of published studies that used administrative data to evaluate the impact of California’s law on provider charges (Erin Duffy’s qualitative research speaks to California providers’ experiences),⁸ and our work differs from previous studies of the New York law in two ways: study sample and study period. Cooper and colleagues examined the impact of New York’s law on ED surprise billing episodes (allowed frequency, allowed amounts, and charges).⁶ We extended this analysis by studying charges in the nonemergency setting. This setting has received less attention and scholarship. In spite of that, we believe that study is warranted for these scenarios because the No Surprises Act covers both emergency and nonemergency surprise bills.

Where provider charges are an input to an arbitrator’s decision, and because charges are set by providers and not insurers, it is possible that over time, providers who might enter arbitration could selectively increase charges for infrequently performed (within a given geographic market) nonemergency procedure codes to receive higher payments during an independent dispute resolution process. As seen in exhibit 3 and appendix section 2,⁷ the gap between out-of-network charges in New York and comparison states seems to have increased over time, which is consistent with providers learning that increasing charges can yield beneficial outcomes in arbitration. We also note that the effect of the law in New York mainly seems to appear approximately one year after enactment. Perhaps this may be due to when providers adjust their charge schedules.

Our research, in comparison with that of Cooper and colleagues,⁶ relied on data extending through 2020. We suspect that in addition to a differing setting (ED versus non-ED), as well as sample composition, one possible reason that we saw rising charges and Cooper and colleagues did not is that our postintervention period was longer, allowing us to pick up potential long-term response to incentives by providers.

It is important to examine provider charges because some state laws do not prohibit arbitrators from considering these amounts when making out-of-network reimbursement determinations. The No Surprises Act permits such states to continue to use arbitration that considers provider charges.⁹ Providers alone are in control of billed charges, and increases to charges can have an impact on payments for surprise bills when charges can be used in independent dispute resolution processes. The long-term success of such a process that relies on charges may be compromised if providers have an incentive to “game the system” by increasing their own payments in disputes via increasing charges. Elevance Health has identified anecdotal situations in which out-of-network providers delivering services that are furnished by few other providers in a given area greatly increased billed charges in a way that skewed the data in the FAIR Health database for a given service (Alison Armstrong, Anthem, Inc., personal communication, June 2021).

When a provider is billing for a procedure that is infrequently performed in a particular geographic area, a single provider’s increase in charges from below to above the eightieth percentile of charges may affect the usual, customary, and reasonable benchmark. Our analysis of provider market share and Herfindahl-Hirschman Indices in appendix section 9⁷ indicates that a substantial number of CPT–ZIP code combinations exist in New York, where a single provider or a small number of providers, especially in the case of assistant surgeons and surgical assistants, would have enough market power to have an impact on the median or eightieth percentile of charges for that service in their area. This suggests the feasibility of strategic provider behavior as a mechanism for the increase in out-of-network charges we observed in New York.

Recent work by Adam Biener and colleagues found that in the ED, physicians who balance bill in surprise billing scenarios collect a greater share of charges compared with other cases (65 percent relative to 52 percent).¹⁰ If the pattern in share of collected charges is similar for nonemergency cases, this could explain why in California we saw a drop in charges relative to comparison states. The removal of the ability to balance bill patients may blunt a financial incentive to retain high charges in California.

Our study shows that after the law was passed in New York, where charges can be used in the arbitration process, there was an increase in charges by out-of-network providers during surprise bill scenarios. Conversely, after the passage of California’s law, which uses a payment standard, there was a decrease in out-of-network charges in California compared with the other states. Narrowing in by specialty, we found that assistant surgeons and surgical assistants had large increases in average charges in New York. Out-of-network assistant surgeons and surgical assistants accounted for a relatively large percentage of surprise bill scenarios, so the increase in charges in this group may have driven the overall increase in out-of-network charges that we observed in New York. These providers may routinely partner with in-network surgeons, and patients typically will choose an in-network surgeon without knowing about the assistant surgeon.

It is important for policy makers to understand the potential consequences of data on billed charges being used during arbitration for surprise bills.

Surprise billing laws that allow for charges to be considered by arbitrators are associated with an increase in provider charges.

It is important for policy makers to understand the potential consequences of data on billed charges being used during arbitration for surprise bills, particularly policy makers in states with laws addressing surprise billing. Although the No Surprises Act prevents arbitrators from considering charges under the federal patient protection, states such as New Jersey still have protections in place that rely on charges as a component of arbitration. Also, although New York has begun the process of aligning its state law with the No Surprises Act,¹¹ it is not known as of the writing of this article whether charges will remain a component of independent dispute resolution processes in practice; more research is necessary. Prior studies showed that awards from arbitration are high, often tethered to the eightieth percentile of charges.^4,5 However, our study further demonstrates that surprise billing laws that allow for charges to be considered by arbitrators are associated with an increase in provider charges, which may ultimately lead to higher costs for out-of-network care.

ACKNOWLEDGMENTS

Benjamin Chartock has received a research grant from Arnold Ventures to study surprise billing. As employees of Elevance Health, Aliza Gordon, Ying Liu, and Winnie Chi hold Elevance Health stocks. They were not compensated for their contributions outside of employment. The authors thank Qian Si, Dianna Hayden, and Ajay Rana, employees of HealthCore, Inc., for programming support.

NOTES

• 1 Garmon C, Chartock B. One in five inpatient emergency department cases may lead to surprise bills. Health Aff (Millwood). 2017;36(1):177–81. Go to the article, Google Scholar
• 2 Cooper Z, Scott Morton F. Out-of-network emergency-physician bills—an unwelcome surprise. N Engl J Med. 2016;375(20):1915–8. Crossref, Medline, Google Scholar
• 3 Chhabra KR, McGuire K, Sheetz KH, Scott JW, Nuliyalu U, Ryan AM. Most patients undergoing ground and air ambulance transportation receive sizable out-of-network bills. Health Aff (Millwood). 2020;39(5):777–82. Go to the article, Google Scholar
• 4 Chartock BL, Adler L, Ly B, Duffy E, Trish E. Arbitration over out-of-network medical bills: evidence from New Jersey payment disputes. Health Aff (Millwood). 2021;40(1):130–7. Go to the article, Google Scholar
• 5 Adler L. Experience with New York’s arbitration process for surprise out-of-network bills. USC-Brookings Schaeffer on Health Policy [blog on the Internet]. 2019 Oct 24 [cited 2022 Jul 18]. Available from: https://www.brookings.edu/blog/usc-brookings-schaeffer-on-health-policy/2019/10/24/experience-with-new-yorks-arbitration-process-for-surprise-out-of-network-bills/ Google Scholar
• 6 Cooper Z, Morton FS, Shekita N. Surprise! Out-of-network billing for emergency care in the United States. J Polit Econ. 2020;128(9):3626–77. Crossref, Google Scholar
• 7 To access the appendix, click on the Details tab of the article online.
• 8 Duffy EL. Influence of out-of-network payment standards on insurer-provider bargaining: California’s experience. Am J Manag Care. 2019;25(8):e243–6. Medline, Google Scholar
• 9 Keith K, Hoadley J, Lucia K. Banning surprise bills: Biden administration issues first rule on the No Surprises Act. Health Affairs Blog [blog on the Internet]. 2021 Jul 6 [cited 2022 Jul 18]. Available from: https://www.healthaffairs.org/do/10.1377/forefront.20210706.903518/ Google Scholar
• 10 Biener AI, Chartock BL, Garmon C, Trish E. Emergency physicians recover a higher share of charges from out-of-network care than from in-network care. Health Aff (Millwood). 2021;40(4):622–8. Go to the article, Google Scholar
• 11 State of New York. Senate–Assembly, budget bill [Internet]. Albany (NY): State of New York; 2022 Jan 19 [cited 2022 Jul 18]. Available from: https://legislation.nysenate.gov/pdf/bills/2021/S8007C Google Scholar